CECL: Not all models are created equal…

Posted by Brad Dahlman

Jun 13, 2018 11:30:00 AM


A lot has been written about the new Current Expected Credit Loss (CECL) changes which will affect FIs in the coming years. This is one of the largest changes to FI financial reporting and credit risk management in decades and certainly warrants ongoing conversations.

Most of the articles written (including mine) have focused on the topics of building a CECL timeline and CECL data elements.  While these topics are often the first steps in the CECL adoption process, I want to start moving the conversation toward some key questions you will face in the near term …


1. Which CECL model(s) are right for my FI?

2. Which models most accurately assess credit risk?

3. What additional capabilities should I seek in a CECL solution that address common challenges?

In the remainder of this blog I will provide some thought and perspective to these questions.

  1. Which model is right for your FI?

To answer this question, you should do your homework by understanding the new requirements and having discussions with interested parties (regulators and accountants). While this is an important step in the process, I doubt you will get clear direction from these sources. The FASB is purposefully vague on these topics and regulators and accountants are likely to fall back on the FASB terminology of “reasonable and supportable.”

Given a lack of clear direction from outside sources, you will need to wrestle internally with the modeling question. Some factors that will affect your CECL modeling decision include:

  • Size of balance sheet.
  • Complexity of balance sheet (aka portfolio concentration/mix).
  • Credit loss history.
  • Internal CECL goals.
    • - “I just want to be … complaint with CECL.”
    • - “I want to be … compliant with CECL and minimize reserves.”
    • - “I want to be … compliant with CECL, minimize reserves, and accurately assess credit risk.”

Organizations that are larger (over $500MM and subject to FDICIA requirements), those with complex balance sheets, and those with higher historical credit losses will likely be required to have more robust models on their adoption date. However, over time I expect regulators and accountants to push for more robust models for even smaller organizations with low credit loss history.

In our highly regulated industry, I can understand the temptation to set the “internal” bar low and strive to simply achieve compliance. However … credit risk is the largest risk facing FIs today and having strong models and credit risk management processes (including early warning signs of credit concerns) is essential in running an FI today. Regardless of your CECL modeling decision, there is going to be a significant cost, effort, and impact on your financials (and operations) associated with CECL. Given that, I would encourage you to strive to find a solution that will accurately assess your credit risk!

  1. Which models most accurately assess credit risk?

In a recent CECL study,(1) several models were compared (using the same data) to determine how effective they were at assessing risk during the turbulent ”great recession.” As you can see in the chart below, some models were significantly more effective than others. This qualitative assessment is critical as you select your model. Finding a model that accurately predicts credit risk is the true goal of CECL! Read the full study to learn more as you assess the effectiveness of various models.

Picture #1

  1. What additional capabilities should I seek in a CECL solution that address common challenges?

At a recent FDIC webinar attended by over 6,000, the FDIC outlined a series of challenges in CECL modeling. As you select your CECL solution, you should also consider how the solution will address these challenges.

Common Challenges for All Loss Rate Methods (FDIC)

Significant adjustments are necessary when:

  1. Losses are minimal
  2. Losses are sporadic with no predictive patterns
  3. There is a low number of loans in each pool
  4. Data is only available for a short historical period
  5. Today’s portfolio composition varies significantly from historical portfolios
  6. There are changes in economic environment (e.g., available historical data is from a recessionary period, but today’s environment is mid-expansionary period)

People #2

How better (or the best, or optimal) CECL solutions address these challenges …

Shared data pool (Challenges 1 – 4) – A shared data pool with a substantial balance and more than a dozen years’ worth of history is optimal. A shared data pool (which is encouraged by FASB) will help address the challenges of minimal/sporadic loss history, low number of loans in a pool, and a historically short period of time. The power of the shared pool and model calibration for your loss history allows you to get a more accurate CECL forecast.

Account level Detail (Challenge 5) – Monthly data extracts of account level data will allow you to identify the changing risk in your portfolios. You will know how much your reserves are impacted by credit improvement (or declines) in your portfolio based on this account-level data compared to prior periods.

Changing Economic Conditions (Challenge 6) – A quarterly forecast from the Federal Reserve Bank that is “regionalized” for your state and automatically fed into your solution will allow you to understand how your FI’s reserves change based on changing economic conditions.

The CECL clock is ticking, and while everyone is challenged to determine the right CECL model(s) and partner, I would encourage you to start thinking now about which model(s) and provider will best help you assess and manage the changing credit risk at your FI.

Topics: Risk Mitigation, Credit Unions, Risk Management, CECL, banking

Subscribe to Email Updates

Untitled Document Untitled Document

Recent Posts